How to Argue Honestly and Still Win

Most advice about winning arguments is advice about cheating: interrupt, reframe, never concede, keep the other person on the defensive. It works in the narrow sense that it can make an opponent look bad in front of an audience. It fails in every sense that matters afterwards. Nobody changes their mind, the audience learns nothing, and the person who "won" has trained themselves to stop noticing when they are wrong.

There is a better tradition. Over the last seventy years philosophers and communication scholars have built a practical account of what a good argument looks like and what the participants owe each other. It is not a call for politeness. It is closer to an engineering standard. People who argue this way tend to win more of the arguments worth winning, because their positions have already survived the strongest objections before they open their mouths.

Frans van Eemeren and Rob Grootendorst, the founders of the pragma-dialectical school, define a critical discussion as an exchange aimed at resolving a difference of opinion on the merits. That definition quietly rules out a great deal. Silencing the other side does not resolve the difference. Neither does exhausting them, embarrassing them, or getting the moderator to move on. On this view you have won when one of two things happens: the other party, or a fair-minded observer, accepts your standpoint because of the reasons given, or you discover that your standpoint was wrong and drop it. The second outcome is not a consolation prize. You leave with a better belief than you arrived with, which is the only lasting thing an argument can give you.

Stephen Toulmin's The Uses of Argument replaced the tidy syllogism of the logic textbook with a layout that fits how people actually reason in law, medicine, and everyday life. A Toulmin-style argument has six parts:

• Claim: what you are asking the listener to accept.
• Grounds: the facts or data you offer in support.
• Warrant: the general principle that licenses the step from grounds to claim.
• Backing: the reason to trust the warrant itself.
• Qualifier: how strongly the claim holds: certainly, probably, in most cases.
• Rebuttal: the conditions under which the claim would not hold.

The warrant is where most real disagreements live, and it is usually unstated. Suppose someone argues that a city should adopt a policy because a neighbouring city adopted it and saw good results. The grounds may be accurate. The hidden warrant is that the two cities are alike in the ways that matter. Say the warrant out loud and the conversation becomes productive immediately: are they alike? In which respects? The honest arguer surfaces their own warrants before the opponent has to dig for them.

The qualifier and rebuttal are the parts that feel like weakness and function as strength. A claim stated as "in most cases, unless X" is far harder to knock down than the same claim stated as a universal law, because the single counterexample that would have sunk the universal version has already been accounted for. Overclaiming is a gift to your opponent.

Daniel Dennett, in Intuition Pumps and Other Tools for Thinking, passes along a procedure he credits to the game theorist Anatol Rapoport. Before criticizing a position you should, in order: restate it so clearly and fairly that its holder would endorse your version; list the points on which you agree, especially any that are not widely shared; mention anything you have learned from the other side; and only then offer your rebuttal.

This is steelmanning with a checklist, and it rests on the older principle of charity. It has three practical payoffs. It protects you from the straw man, which is often committed by accident rather than malice, because we tend to remember the weakest version of views we dislike. It makes the other person far more willing to listen, since they have just heard evidence that you did. And it tests your own understanding. Research on the illusion of explanatory depth by Philip Fernbach and colleagues found that asking people to explain step by step how a policy they favoured would actually work tended to reduce their confidence and moderate their stated positions, whereas merely asking them to list reasons for their view did not. If you cannot restate the opposing case, you may not understand the issue as well as you feel you do.

Van Eemeren and Grootendorst set out ten rules for critical discussion and treat a fallacy as any move that breaks one of them. A few carry most of the weight in everyday disputes.

• Freedom. Neither party may prevent the other from advancing or questioning a standpoint. Threats, mockery of the person, and appeals to pity designed to make a challenge feel cruel all violate it. The ad hominem is, on this analysis, an attempt to disqualify a speaker rather than answer them.
• Burden of proof. Whoever advances a standpoint must defend it when asked. "Prove me wrong" is burden shifting. See also the entry on how the burden is properly allocated.
• Standpoint. An attack must address the position the other party actually took. This is the rule the straw man breaks, and the one the motte-and-bailey exploits from the other direction, by swapping a bold claim for a modest one when challenged.
• Relevance. A standpoint may be defended only with argument that bears on it. Whataboutism changes the subject to the critic's own side, which may be a fair topic for another day but is not a defence.
• Closure. A failed defence means retracting the standpoint; a successful defence means the challenger withdraws their doubt. Almost nobody does this in public, which is precisely why doing it is so persuasive to onlookers.

The remaining rules cover unexpressed premises, agreed starting points, the soundness of argument schemes, logical validity, and clarity of language. You do not need to memorize them. The useful habit is to ask, when a discussion goes bad, "which obligation was just skipped?"

Douglas Walton's work is a corrective to the habit of shouting fallacy names at people. In Informal Logic he argues that most of the classic fallacies are reasonable forms of argument that become fallacious only when misused in context. Appealing to an expert is not an error; we could not function without it. It becomes one when the critical questions attached to that scheme go unanswered. Is this person actually an expert in the relevant field? Did they really say this? Is it consistent with what other experts say? Do they have a stake in the answer? Is the opinion backed by evidence that could be examined?

This gives the honest arguer something better than a label. Instead of announcing "appeal to authority," ask the critical question that the argument has failed to meet. The label invites a quarrel about definitions. The question invites an answer, and if no answer comes, the audience notices.

Arguing honestly does not mean being defenceless. Some practical moves:

• Against a Gish gallop, do not chase every point. Say that a dozen claims were made in a minute, pick the one the speaker seemed to rely on most, answer it thoroughly, and note that the rest can each receive the same treatment.
• Against a shifting position, write the claim down. "Let me make sure I have it: you are claiming X?" A standpoint stated in plain words is much harder to swap later.
• Against a personal attack, name the move once, without heat, and return to the question. The audience is judging composure as well as content.
• Concede what is true. A concession costs you a point and buys you credibility on every other point. Refusing to concede the obvious has the reverse exchange rate.
• Know when it is not a critical discussion. If the other party has no intention of resolving anything, your real audience is whoever is watching, and sometimes the right move is to stop.

Before an argument that matters, write four lines: your claim with its qualifier; your best grounds; the warrant that connects them; and the strongest objection, stated as its smartest advocate would put it. If you cannot fill in the fourth line, you are not ready. If you can, and your claim still stands, you will find that you rarely need any tricks.

For the dishonest counterparts to everything above, browse the logical fallacies in the frameworks section, or run a transcript through the detector to see which moves it flags.